OpenAI’s newly released GPT-6.1 Sol has landed just one point behind the company’s flagship GPT-6 Astra on the Artificial Analysis Intelligence Index, according to the independent benchmarking firm. GPT-6.1 Sol, running at max reasoning effort, scores 52 on the index, against 53 for GPT-6 Astra at max effort.
That is a significant jump from GPT-6 Sol, which scored 48 on the same index at max effort when it launched last week. The four-point gain in a matter of days puts the upgraded model level with Astra’s standing on the benchmark for a fraction of the price: OpenAI lists GPT-6.1 Sol at $2 per million input tokens and $10 per million output tokens, one-fifth of Astra’s $10 and $50.
The result lines up with OpenAI’s own claim that GPT-6.1 Sol nearly matches Astra’s intelligence on agentic coding, computer use and professional work.
Where the rest of the field stands #
The latest version of the index (v4.3.2) combines ten evaluations: AA-Briefcase v1.1, GDPval-AA v2.1, AutomationBench-AA, Terminal-Bench 4.0, SciCode, Humanity’s Last Exam, GDP.pdf, CritPt, AA-Omniscience and AA-LCR v1.1.
On that index, Anthropic’s models still hold the top of the chart. Claude Opus 5.5 leads with a score of 58, followed by Claude Sonnet 5.5 at 56. Claude Fable 5.1 scores 53, tying GPT-6 Astra. All three Anthropic scores are for the max setting with fallbacks enabled.
GPT-6.1 Sol sits just below that group at 52, and ahead of Claude Opus 5 (max) at 51. Below it, Meta’s Muse Spark 1.3 (max) and GPT-6 Sol (max) are tied at 48, followed by xAI’s Grok 4.7 (xhigh) and Xiaomi’s MiMo-V2.6-Pro at 46. Alibaba’s Qwen3.8 Max and Zhipu’s GLM-5.3 (max) score 45 each, while Step 5 Preview and Moonshot’s Kimi K3 (max) are at 44. GLM-5.3-Flash scores 42 and Google’s Gemini 3.8 Flash (high) 41.
Price is the story #
The headline for GPT-6.1 Sol is cost. OpenAI has kept the model’s standard API pricing at $2 and $10 per million input and output tokens, the same as GPT-6 Sol, and has halved the cached input price to $0.10 per million tokens. That makes it a much cheaper way to get close to Astra-level performance, particularly for agents that reuse a lot of context across requests.
Artificial Analysis has not yet published a cost-per-task figure for GPT-6.1 Sol in the chart shared with the result. For reference, GPT-6 Sol at max effort cost about $1.06 per task to run the index, so the new model’s total cost will depend on how many tokens it uses to reach its higher score.
GPT-6.1 Sol is available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users, and through the API as gpt-6.1-sol. It is not yet available in Chat.